Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [] A suggested topic for PlanningCouncil Discussion

I think it's easy to agree that overall quality is vital to Eclipse, but I
have to wonder if the release train is the right way to enforce it. For the
sake of openness, transparency, inclusion, etc.,  Eclipse projects are very
public about on-going progress, and there is a community vote as part of
the release. If someone documented serious quality issues via the standard
channels and enough of the community agreed with the assessment, then its
pretty obvious what would happen (I assume the EMO would require the
project to address the problems before release).

Given this (as-designed) substantial community influence, then where is the
prerogative of the Planning Council for making up some number of rules
around a simultaneous release? Are we saying that these rules are more
important than the community consensus? (we're all part of this community,
so this is an extension of Ed's peer-pressure argument to a larger base).

Regarding the ability to scale processing of  IP requests to larger release
trains, I'm of two minds here. Its nice to be on the release train and
thereby have assurance that your IP issues will be processed. :-) But part
me also feels that it is unfair to projects not on the release train.....

John Graham
Eclipse Data Tools Platform PMC Chair
Staff Software Engineer, Sybase, Inc.

             <mike.milinkovich                                          To 
   >             "''"    
             Sent by:                  < 
             ing-council-bounc                                          cc 
                                       RE: []  
             11/01/2007 03:55          A suggested topic for               
             PM                        PlanningCouncil Discussion          
             Please respond to                                             
             Please respond to                                             

Doug, Doug, Ed, et al,

What you are suggesting is an even lower bar than what we have had in the
past. At least on paper, if not in practice.

The problem with this approach is it means that the release trains just get
bigger and bigger and with no incremental improvement in the overall
quality of what’s coming from Eclipse. Shipping a big bag of stuff that
doesn’t work together is not going to help us build a reputation for
quality. It will destroy it. And once you have destroyed a reputation for
quality, it can take a generation (e.g. ~20 years) to get it back, if ever.

As a purely practical matter, I honestly doubt that the Eclipse Foundation
as the IP resources to review and approve all the CQs to ship 30 projects
on the same day. So if you guys don’t come up with some rules that raise
the bar and limit who has the process maturity and quality to get in, don’t
get mad at me for making rude and arbitrary decisions J

I completely understand that what you’re recommending is the simplest and
easiest approach. But IMHO (a) it’s the wrong thing to do for the Eclipse
community and (b) it is unlikely to work in practice.

Mike Milinkovich
Office: +1.613.224.9461 x228
Mobile: +1.613.220.3223

[] On Behalf Of
Gaff, Doug
Sent: Thursday, November 01, 2007 2:32 PM
Subject: RE: [] A suggested topic for
PlanningCouncil Discussion


As far as I’m concerned, the only reason to kick a project off the train is
if they consistently fail to build and update their site at each milestone.
Simply put, the ejection is because “Project X keeps holding up the
release.”  Furthermore, I think it should come to a vote by all of the
projects on the train to kick a single project off.

Everything else should be a strongly encouraged optional requirement, and
we should use public humiliation to police those requirements, e.g. “I
noticed that Project Y is not optimizing their jars, shame on you.  Please
fix it.”  Clearly there are technical must do’s that physically put a
project on the train, and they should be stated as such.

Bottom line, I think we should err on the side of inclusion, and leave it
up the projects to prove that they can or can’t keep up with the milestone
schedule.  If they can’t keep up, then their processes aren’t mature

Doug G

[] On Behalf Of
Scott Lewis
Sent: Wednesday, October 31, 2007 5:17 PM
Subject: Re: [] A suggested topic for
PlanningCouncil Discussion

Hi Bjorn,

Bjorn Freeman-Benson wrote:
Doug, (and everyone)
I agree - if there are no people or people hours, there will be no code, no
matter how much the Board wishes for it to happen. One could argue (I have
argued) that the Board controls the people hours, so if they want to define
a requirement, they should supply the resources, but somehow that logical
situation doesn't always come true.

Do you really think it would poison the community if there were a two-level

I think it would poison the community to have a two-level train.  I think
we would quickly see different requirements and EMO treatment (and member
company support) for the 'corporate-run' projects relative to all the other
projects...those led by smaller companies and/or independents.  Seems to me
this would eventually lead to inequities that many committers would
consider unacceptable for a merit-and-value-based community.

A "meet all the requirements" level (the gold medal) and a "simultaneously
release" level (the silver medal)? Maybe if the packages and the main
update site contained the gold seal projects, but that the silver projects
were also (if there was time to review the IP) available at the same time?

It seems to me like this sort of classification would be inherently
detrimental to 'silver medal' projects and differential to 'gold medal'
projects.  That is, it may say nothing about their usefulness, and/or value
to be labeled as 'silver', but just the labeling by the membership and
foundation will lead to end-user biases...with lower adoption, tougher
distribution, etc., etc.

It does seem to me that if the Board wants to mandate that the projects
have to do more/other in terms of integration/testing, etc for the release
train...and that the EMO should/must police/enforce the new rules...that
there should be some recognition that this implies some support from the
membership to do the work involved.  There are many ways that I can think
of to do this (contributing integration testing resources, allowing
existing committers to work on related projects, etc., etc).    Unfunded
mandates don't really work IMHO...either for the committer community or for
the EMO.


- Bjorn

Doug Schaefer wrote:
As for requirements, other than holding up the IP process I’m not sure what
stick the EMO has to enforce projects meet the requirements. If projects
don’t have the resources or the mandate from the employers of the resources
to do the work, it doesn’t happen. And if you kick projects off the train
because of that, that could poison the community. The best stick still is
influencing and that involves good communication channels open between the
requirers and requirees, and, of course, a reasonable set of requirements.
[end of message]

_______________________________________________ mailing list

 _______________________________________________ mailing list

Back to the top